Introduction

Pseudomonas aeruginosa is a gram-negative pathogen that causes opportunistic infections in susceptible hosts. It is a leading cause of acute pneumonia in hospitalized patients and is responsible for chronic lung infection in patients with cystic fibrosis. Its ability to cause both acute and chronic infections can be attributed to its broad arsenal of virulence factors. Specifically, the type III secretion system (TTSS) has been shown to be a major virulence determinant in the pathogenesis of acute infections.

P. aeruginosa utilizes the TTSS to deliver effector toxins (ExoS, ExoU, ExoY, and ExoT) directly into host cells, which can cause rapid cell necrosis or can modulate the actin cytoskeleton, allowing the pathogen to invade host cells and evade phagocytosis [1]. Depending on the disease site or background, genes encoding the cytotoxins ExoU and ExoS are present as variable traits and are mutually exclusive in most strains, the exoS genotype being the more prevalent of the two. The exoS genotype accounted for 72% of the 115 clinical and environmental isolates examined in one study while 28% of the strains contained the exoU gene. Specifically, a much lower proportion of P. aeruginosa obtained from the respiratory tract was comprised of exoU+ strains than those from blood or wound (18% vs 40%, respectively) [2]. In a murine model of acute P. aeruginosa pneumonia, ExoU has been demonstrated to have the greatest impact on virulence relative to the other TTSS effector proteins (ExoS and ExoT) by measurement of mortality, bacterial persistence in the lung, and dissemination [4]. These experimental findings are supported by clinical studies where ExoU-secreting strains have been associated with poor outcomes of pneumonia [5] as well as persistence and severity of disease [6], [7]. More recently, El-Solh et al. [3] reported on the increased 30-day mortality of bacteremic patients infected with strains expressing the TTSS compared to those infected with non-TTSS expressing strains. Notably, none of the patients infected with an ExoU-secreting strain survived past 30 days.

The emergence of antibiotic-resistant P. aeruginosa strains has presented significant therapeutic challenges. The fluoroquinolone (FQ) antibiotics, ciprofloxacin and levofloxacin, exhibit potent in vitro anti-pseudomonal activity; however, due to its widespread use over the past decade as the most commonly prescribed antibiotic class to adults in the U.S., resistance has developed in parallel to its popularity in use [8], [9]. Of concern, a significant portion of the FQ-resistant strains are also multidrug-resistant [10], [11].

P. aeruginosa acquires resistance to the fluoroquinolones mainly through mutations in the quinolone resistance determining region (QRDR) of target genes essential for the topological maintenance of the genome, DNA gyrase and topoisomerase IV [12]–[14]. Both enzymes are composed of two subunits, encoded by the genes gyrA, gyrB, and parC, parE, respectively. DNA gyrase is responsible for catalyzing the negative supercoiling of DNA, and topoisomerase IV decatenates daughter replicons. Both functions are essential for successful replication of the genome. The reaction cycle includes breakage and resealing of double-stranded DNA. Fluoroquinolones bind to the enzymes and prevent the completion of the reaction, resulting in stalled replication forks and double stranded DNA breaks [15]. Mutations in the four target genes that confer resistance to the fluoroquinolones have been well described in P. aeruginosa [16], [17], with GyrA as the primary fluoroquinolone target [14], [18]. Other mechanisms of resistance include overexpression of efflux pumps as well as innate impermeability of the cell wall [19].

Previous work in our lab has shown that patients infected with FQ-resistant P. aeruginosa strains have 3-fold higher mortality and prolonged illness compared to those infected by susceptible strains [10]. In addition, we reported a significant association between FQ-resistance and the exoU TTSS genotype in 45 clinical isolates of P. aeruginosa obtained from various body sites exhibiting a range of susceptibilities to fluoroquinolones, suggesting a selection bias for development of FQ resistance based on TTSS effector genotype of the strains [20]. A recent study by Garey et al. analyzed bloodstream isolates of P. aeruginosa and found that exoU+ strains were more frequently multi-drug resistant compared to exoS+ strains [21]. Additional analysis on the data indicated that the highest frequency of resistance among the exoU+ strains was towards the fluoroquinolones (87%, 13/15 vs 37%, 7/19 respectively, p = 0.0034). (personal communication) Similar results were observed by others examining corneal isolates [22]. Interestingly, Fleiszig et al has shown that resistance of P. aeruginosa to contact lens disinfectant was linked to acute cytotoxic activity from ExoU towards corneal epithelial cells and exsA, a transcriptional regulator of genes encoding effectors of the TTSS [23]. We hypothesized that P. aeruginosa strains, depending on their TTSS effector genotypes, differentially develop resistance-conferring mutations resulting in a selection bias for a FQ-resistant population predominated by exoU+ strains. Given the impact of ExoU on the outcomes of experimental and human lung infections, we tested our hypothesis in a large sample of clinical isolates obtained from respiratory sources; most of the patients from whom the isolates were obtained had pneumonia. We compared the following by TTSS effector genotype (exoU vs exoS): 1) prevalence and degree of FQ-resistance, 2) efflux pump overexpressed phenotype, and 3) mutations in target genes conferring FQ resistance.

Materials and Methods

Susceptibility Testing and Sequencing for Target Site Mutations

Bacterial isolates were stored in cryovials at −80C until ready for testing. Clonality of the clinical isolates was assessed by random amplified polymorphic DNA polymerase chain reaction (RAPD-PCR) using a primer sequence published previously [24] (Table 1). Susceptibility testing to levofloxacin was performed by broth microdilution in 2-fold dilution at concentrations ranging from 0.25 to 128 µg/ml according to guidelines recommended by CLSI [25]. FQ resistance is defined as levofloxacin MIC ≥4 µg/ml. Both the TTSS effector genotype (exoU and exoS) [20], [26] and the QRDR region of respective target genes, gyrA, gyrB, parC and parE was assayed by polymerase chain reaction (PCR) using previously published primers and conditions [17], [27]. The latter was sequenced to identify active and silent mutations compared to wild-type strains PAO1 (exoS+/exoT+/exoY+) and PA103 (exoU+/exoT+/exoY+). Primer sequences are listed in Table 1. Efflux pump overexpressed phenotype was determined by use of a commercially purchased efflux pump inhibitor (EPI, MC-0228) (Sigma) previously shown to broadly inhibit the multidrug Mex efflux pumps of P. aeruginosa in which the fluoroquinolones are substrates [10], [28]. An overexpressed phenotype is defined as ≥8-fold decrease in levofloxacin MIC when tested in the presence of EPI at 20 µg/ml previously established based on the change in MICs between wild type and efflux pump knockout mutants [28].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Primers used for PCR and sequencing.

https://doi.org/10.1371/journal.pone.0042973.t001

Results

Bacterial Strains

A total of 270 P. aeruginosa isolates obtained from respiratory specimens from different patients were evaluated in this study. Majority of the strains (93%) caused pneumonia while the remainder caused bronchitis or colonized the airways. About 20% of the patients with pneumonia had ventilator-associated pneumonia. Clonality was determined for most strains (n = 223); those strains carrying both or neither genes encoding ExoU and/or ExoS were excluded. Strains with identical band patterns were considered to be related and assigned a unique random amplified polymorphic type. The 63 FQ-resistant exoU strains assessed fell into 34 clonal groups, and the 27 FQ-sensitive exoU strains assessed fell into 17 clonal groups. Among the exoS isolates, the 65 strains analyzed fell into 37 groups, and the 68 FQ-sensitive strains analyzed are in 26 groups. Overall, there are no more than 9 isolates in any given clonal group (Table 2). Clonal distribution was similar between FQ-resistant and FQ-sensitive strains with 51% (65/126) and 45% (43/95) of strains belonging to unique RAPD types, respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. RAPD Results.

https://doi.org/10.1371/journal.pone.0042973.t002

Prevalence and Degree of FQ resistance

Overall, less than half of the 270 isolates (38%, 103/270) were exoU+ while 56% (152/270) were exoS+; five isolates were positive for both effector genes (2%) while 10 (6%) had neither. Similar proportion of exoU+ and exoS+ strains caused pneumonia (85%, 87% respectively). Slightly over half (54%, 145) of the isolates were FQ-resistant. The distribution of MIC values was similar between exoU+ and exoS+ populations. MIC50 and MIC90 of all 270 clinical strains were 8 and 64 µg/ml, respectively. Notably, a significantly higher proportion of exoU+ strains were FQ-resistant compared to exoS+ strains (63%, 65/103 vs 49%, 75/152; p = 0.03). (Figure 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Frequency of fluoroquinolone-resistance among exoU vs. exoS containing isolates.

Fluoroquinolone (FQ) resistance is defined as levofloxacin MIC ≥4 ug/ml. FQ-susceptible population is predominated by exoS+ isolates while FQ-resistant population is predominated by exoU+ isolates. *p = 0.03.

https://doi.org/10.1371/journal.pone.0042973.g001

Target Site Mutations

The QRDR of the 4 target site genes (gyrA, gyrB, parC, and parE) was amplified by PCR and sequenced for resistance-conferring target site mutations (TSMs) in a subset of 108 isolates; 59 (55%) of which were exoU+ and 49 (45%) were exoS+ isolates. Table 3 summarizes all 108 isolates that have been sequenced for target site mutations. In general, MIC increased as the number of active mutations increased. Out of all the isolates with at least one target site mutation, only 4 had levofloxacin MIC <4 µg/ml. A total of 5 exoU+ and 8 exoS+ strains had no target site mutations; all showed the EPO phenotype. Among the 91 isolates with active mutations, a trend towards more single TSM was observed in exoS+ strains (54%, 22/41 vs 38%, 19/50; p = 0.135) while more exoU+ strains had combined TSMs (62%, 31/50 vs 46%, 19/41; p = 0.135). Interestingly, exoU+ strains were more likely to acquire 2 or more TSMs than exoS+ strains at lower MICs (≤8 µg/ml), 13% (4/31) vs. none. (Figure 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Cumulative Frequency of Isolates with ≥2 target site mutations at each given MIC comparing exoU- vs exoS-containing isolates.

The proportion of isolates that has acquired at least two target site mutations is greater with exoU+ than exoS+ isolates and increases in a linear fashion as MIC increases, with exoU+ isolates starting at lower MICs. At a levofloxacin MIC ≤8 ug/ml, 25% of exoU+ isolates compared to 0% of exoS+ isolates have 2 or more TSMs whereas for isolates with MICs ranging from 16–32 ug/ml, 36% exoU+ isolates compared to 22% of exoS+ isolates, and for isolates with MICs ranging from 64–128 ug/ml, 52% exoU+ isolates compared to 39% of exoS+ isolates have 2 or more TSMs.

https://doi.org/10.1371/journal.pone.0042973.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Comparison of target site mutations (TSMs) by type III secretion effector genotype.

https://doi.org/10.1371/journal.pone.0042973.t003

Substitution of isoleucine for threonine at position 83 in gyrA was the most commonly noted mutation among those with active mutations (88%, 80/91) and in similar proportion between exoU+ and exoS+ isolates (90%, 45/50 vs. 85%, 35/41, respectively). Of those with only this mutation at position 83 of gyrA, all except one had levofloxacin MIC ≥8 µg/ml. Even after addition of the EPI, about 37% (11/30) of those isolates remain resistant, with MIC ranging from 4 to 16 µg/ml. A total of 5 isolates have a different gyrA amino acid substitution, asparagine for aspartic acid at position 87. This mutation contributed only moderately to resistance in strains with combined TSM and EPO phenotype; levofloxacin MICs ranged from 1–16 µg/ml in those isolates, which lowered to ≤0.5 µg/ml after addition of EPI. In total, only 2 isolates each had any active mutations in gyrB, parC, or parE without a concurrent gyrA mutation; the corresponding levofloxacin MICs after addition of EPI for those isolates were: gyrB (2, 0.25 µg/ml), parC (8 µg/ml in both), and parE (2 µg/ml in both).

Notable differences in specific TSMs (single or combined) were observed between exoU+ and exoS+ isolates. Mutations unique to the exoU+ strains included a double amino acid change in GyrA at positions 83 and 87, a single substitution at position 468 in GyrB, and a triple mutant with two GyrA changes (positions 83 and 87) and a ParC substitution at position 87. Mutations unique to the exoS+ isolates result in a change from lysine to arginine at position 65 of GyrA, and a double gyrA mutant similar to the one found in the exoU+ strains at positions 83 and 87. One exoS+ isolate also carried a unique mutation in gyrB which resulted in a substitution of tyrosine for serine at position 466. Unlike any exoU+ isolate in this study, one exoS+ isolate had a single mutation in parE only, which resulted in a levofloxacin MIC of 16 µg/ml and 2 µg/ml in the presence of EPI.

Interestingly, among isolates with combined TSMs, mutation in gyrA is present in all isolates, but the proportion of exoU+ and exoS+ strains differ significantly in the subunit of topoisomerase IV affected. (Table 2) Strains containing the exoU gene were twice more likely to have combined mutations in both gyrA and parC than exoS+ strains (48%, 28/59 vs. 24%, 12/49, p = 0.0439, OR 2.369 (95%CI: 1.014–5.536). On the other hand, exoS+ strains were more likely to have combined mutations in gyrA and parE.

Many silent mutations were also found in the target genes; gyrB and parE were the most affected, with 84% of all the isolates carrying one or more silent mutations in gyrB and 86% carrying one or more in parE. Silent mutations occur more frequently in the gyrA gene among exoS+ than exoU+ isolates (74%, 36/49 vs. 20%, 12/59; p<.0001). On the contrary, significantly more exoU+ isolates have silent mutations in the parE gene than exoS+ isolates (93%, 55/59 vs. 78%, 38/49. p = 0.019). The presence or number of silent mutations did not correlate with levofloxacin MIC.

Discussion

Clinical isolates of P. aeruginosa carrying the gene encoding for either the ExoU or the ExoS TTSS effector proteins have been documented to differentially impact virulence in pneumonia as well as in the development of antimicrobial resistance. Compared to exoS+ strains, exoU+ strains have been shown to have higher cytotoxicity correlated with increased resistance, in both in vitro [20], [23] as well as in vivo murine models of infection [6], [29]. In a study of population structure P. aeruginosa isolates in Europe, the exoU+ genotype was found to be significantly associated with multidrug resistance and ciprofloxacin resistance compared to the exoS+ genotype [30]. The present study sought to explore the observed link between TTSS virulence and fluoroquinolone resistance in P. aeruginosa. We showed that exoU+ and exoS+ strains differentially acquire resistance-conferring mutations and that exoU+ strains may be genetically favored in a fluoroquinolone-rich environment that stemmed from heavy prescribing.

P. aeruginosa develops resistance to the fluoroquinolones primarily from acquiring target site mutations or overexpression of multidrug efflux pumps from the Resistance-Nodulation-Division (RND) family [31]. Confirming our previously published findings, we have shown here that the fluoroquinolone-resistant subpopulation is predominated by exoU+ strains in a large sample of respiratory isolates, despite the higher prevalence of exoS+ strains overall. This may be related to our observation that exoU+ strains more readily acquire multiple target site mutations and at a lower MIC than exoS+ strains. Overexpression of multidrug efflux pumps belonging to the RND family of proteins results from mutations in genes that encode for the pumps or in genes that regulate their expression [31]. Here, we utilized an efflux pump inhibitor to assess the extent of efflux pump involvement in resistance by comparing the MICs of levofloxacin before and after the addition of efflux pump inhibitor. Efflux pump-overexpression significantly contributed to FQ-resistance in both exoU and exoS-containing isolates.

With respect to target site mutations, those observed in our isolates are consistent with previously published mutations [13], [14], [17], [27], [32]. Substitution of isoleucine for threonine at position 83 in GyrA is the most frequent TSM in our cohort of FQ-resistant isolates. Higher MICs are associated with an accumulation of mutations, the most common being a double mutation in gyrA and parC. Others have found strains with a double gyrA-parC mutation to be 3–4 times more resistant to ciprofloxacin than strains with a single mutation in gyrA [12]. Similar results were found in Salmonella sp. where mutations in gyrA and parC as well as double mutations in gyrA conferred high level resistance [33]. Our results provide further support for this stepwise accumulation of mutations mechanism [27]; only 2 resistant isolates have single mutations in parC without the “first step” threonine 83 mutation in gyrA. This shows that although GyrA is the main target of fluoroquinolones, a mutation in gyrA is not a strict prerequisite for the acquisition of mutations at other target genes leading to resistance. Despite previous results describing single mutations in the gyrB and parE genes not conferring much resistance [32], we found relatively high levofloxacin MICs ranging from 8–32 µg/ml (0.25–2 µg/ml in the presence of an EPI) in 2 isolates with a single gyrB mutation and one with a single parE mutation.

One striking difference observed between TTSS genotype in FQ resistance relates to the presence of specific active target site mutations conferring resistance. Isolates containing the exoU gene were twice as likely to have mutations in both gyrA and parC than exoS+ isolates. Furthermore, the type of mutations in the parC gene differed with respect to TTSS effector genotype. The serine to leucine change at amino acid 87 in ParC was more likely to occur in exoU+ strains, while the glutamate to lysine change at position 91 was more likely to occur in exoS+ strains. This observed difference may point to a differential adaptive response of exoU+ and exoS+ to fluoroquinolone exposure.

Synonymous mutations have also been shown to have an effect on gene expression. When a library of green fluorescent protein genes that differed only in codon usage but not in the amino acid sequence was expressed, the levels of expression varied among the polymorphic genes, indicating that cellular processes can be affected without changing the amino acid sequence [34]. This is likely a result of a change in the mRNA secondary structure. Interestingly, we observed in this study differences in the prevalence and type of silent mutations that occurred in the QRDRs of the four target genes by TTSS effector genotype. The gyrB and parE genes had the highest number of silent mutations possibly because they are not the main contributors to enzymatic action and therefore can tolerate the polymorphisms. exoS-containing strains have a significantly higher rate of silent mutations in gyrA than exoU+ strains while the opposite is observed for the exoU+ strains with the parE gene; both TTSS effector genotypes have similar rates of silent mutations in gyrB and parC.

The results presented in this study show that resistance-conferring mutations developed differently in exoU+ and exoS+ strains. We speculate that the mechanism for this difference may involve the SOS response, a global response that allows bacteria to tolerate sudden increases in DNA damage. Fluoroquinolones induce the SOS response by poisoning the topoisomerase enzymes, leading to double stranded DNA breaks. Sub-MIC levels of ciprofloxacin have been shown to rapidly induce the SOS response in S. typhimurium [35]. A subset of genes induced by the SOS response include the error prone polymerases (Pol II, Pol IV, and PolV) [36]. These low-fidelity polymerases are able to perform translesion synthesis, enabling stalled replication forks to continue past DNA damage, but at the expense of introducing errors into the genome. By preventing induction of the SOS response in vivo and in vitro, pathogenic E. coli were unable to acquire resistance mutations [37]. This same study also showed that if any of the SOS-induced polymerases were made non-functional, the mutation rate was much less than in wild type. Relating those results to our findings leads us to question whether a difference in the SOS response in exoU+ and exoS+ strains could account for the differences in target site mutations observed. It is possible that differential quinolone exposure in patients from whom the strains were obtained may contribute to the difference in mutational processes between the two TTSS effector genotypes. We have minimized this potential bias by focusing only on respiratory isolates that caused acute infections in this study.

Differences in the target genes affected by resistance-conferring mutations in exoU+ vs. exoS+ strains may have important biological implications related to virulence expression. Promoters of certain genes have been shown to be especially sensitive to changes in the level of supercoiling, affecting expression levels. It has been proposed that supercoiling alters the structure of the promoter region, determining its accessibility to the transcription machinery [38]. Altered levels of supercoiling via inhibition of gyrase were shown to alter the expression levels of virulence genes in Salmonella typhimurium [39] as well as Staphylococcus aureus [40], suggesting that transcription of virulence genes may be regulated through changes in supercoiling. Others have shown that controlling the topology of the genome can be a mechanism of global gene regulation. An analysis of expression levels after exposure to a gyrase inhibitor in Streptococcus pneumoniae showed the genome to be organized in topology-reacting gene clusters [41]. The levels of supercoiling in fluoroquinolone-resistant E. coli strains with various target site mutations have been shown to differ [42] while altered protein abundances were observed among strains of E. coli with mutations in the gyrase gene compared to wild type [43]. Taken together, it is tempting to speculate that the different target site mutations in exoU+ and exoS+ strains could potentially affect topoisomerase activity and lead to differential expression of virulence genes with topology-sensitive promoters.

We present here evidence of differentiation in quinolone resistance by virulence genotype in a large sample of respiratory isolates of P. aeruginosa. A limitation of the present study is a lack of data regarding prior antibiotic exposure. The retrospective nature of the clinical component of the study is limited by the inconsistent documentation of antibiotic exposure in patients' medical charts. It is possible that prior antibiotic exposure in the affected patients drove the development of resistance in these strains during the course of infection; alternatively, acquisition of these resistant strains causing acute infections in these patients may be through horizontal transmission. Unpublished data from our lab testing exoU+ (PA103 and one clinical isolate) and exoS+ strains (PAO1 and one clinical isolate) serially passaged under increasing FQ concentrations indicates that the target site mutations accumulated in the resistant mutants in vitro were primarily GyrA mutation and do not mimic those that were observed in the large sample of clinical isolates that we surveyed in this study. The discordance in resistance development between in vivo and in vitro conditions likely reflects the complex in vivo environment that impacts the mutational process in ways beyond drug exposure. Future experiments using isogenic strains are planned to confirm whether the selection for resistant exoU+ subpopulations imposed by heavy fluoroquinolone use is a form of genotype-by-environment interaction to ensure fitness of Pseudomonas evolving in a quinolone-rich environment. Importantly, negative outcomes reported in the literature in association with fluoroquinolone resistance in Pseudomonas infections may be attributable in part to the shift in population dynamics in which resistance and virulence traits co-evolved, favoring the more virulent exoU+ strains. The current practice of widespread fluoroquinolone prescribing has far reaching negative consequences beyond antimicrobial resistance.

Acknowledgments

We thank Samuel Lee for his technical contribution to the PCR experiments, Christine Vu for performing RAPD assays, and Steven Finkel, Ph.D. for his critical review of the manuscript and helpful comments and suggestions. This study was supported by Award Number R21AI073467 from the National Institute of Allergy and Infectious Diseases to AWB.